Geologic Setting and Age

The holotype and only known specimen of the new taxon (USNM PAL 768729) consists of associated elements that were recovered from freshwater limestone collected near University of Michigan fossil locality SC-4, the 8abc limestone (fig. 1; see Beard and Houde, 1989; Bloch and Bowen, 2001; Bloch and Boyer, 2001; Bowen and Bloch, 2002). The fossil bed is early Eocene in age (Wa-1 faunal zone of the Wasatchian North American Land Mammal Age; 1570 m above the K-Pg boundary), in the Willwood Formation, Clarks Fork Basin, Park County, Wyoming. Detailed locality information is archived at the University of Michigan Museum of Paleontology (see also Gingerich, 2001). We take its age to be 56–55 Ma.

Size

To estimate snout-vent length (SVL), we created a dataset of tooth height (measured in the middle of the maxilla) and SVL for iguanid lizards (appendix). Although the taxon sampling might not be ideal, the clade Iguanidae (= Pleurodonta) is close to Anguimorpha in molecular phylogenies, neither Iguanidae nor Xenosaurus shows a marked propensity to lengthening of the teeth, and insufficient measurable specimens of Anguimorpha were available. Moreover, owing to the paucity of small anguimorph species, extrapolation would have been necessary. Because taxa classified as Xenosaurus generally have more vertebrae than typical iguanids (Gauthier et al., 2012), our estimates of SVL for USNM PAL 768729 and other stem xenosaurs are likely to be slight underestimates. We ln-transformed the data, regressed ln(SVL) on ln(tooth height), estimated ln(SVL) and 95% confidence interval on the slope, and then back-transformed ln(SVL) to obtain SVL.

Phylogenetic Analyses

Maximum parsimony: To infer the phylogenetic relationships of the new taxon, we coded morphology of the holotype using the 274-character data matrix of Bhullar (2011), which, while focused on xenosaurs, includes representatives of all major anguimorph lineages. We excluded Carusia intermedia, as this species has recently been determined to be a stem scincid (Gauthier et al., 2012; Reeder et al., 2015) as well as the fossil helodermatids Eurheloderma gallicum and Primaderma nessovi. The iguanid Pristidactylus torquatus was used as the outgroup. Thus, the modified dataset included 24 species. We added the following new characters [appended to the end of the Bhullar (2011) matrix so character numbers are unchanged]:

Of the 280 characters, 256 are parsimony informative. The modified matrix was analyzed (1a) without any topological constraint using maximum parsimony (MP) in PAUP* v.4.0a release 161 ( https://paup.phylosolutions.com) with the following settings: heuristic search, random addition sequence, 1000 repetitions, TBR branch-swapping. Because molecular data strongly support a different topology of anguimorph phylogeny than the morphological one (e.g., Fry et al., 2006; Pyron et al., 2013), we conducted a second analysis (1b) in which a minimum basic outline of the molecular tree (Reeder et al., 2015) was enforced as a backbone: (((Heloderma suspectum (Elgaria multicarinata, Xenosaurus grandis)), (Shinisaurus crocodilurus (Lanthanotus borneensis, Varanus exanthematicus))). In each analysis, clade support was assessed using 1000 bootstrap pseudoreplicates, each with the same settings given previously.

A posteriori time scaling: The single MP tree was time-scaled a posteriori using the three-parameter calibration model of Bapst (2013). The three rate priors (speciation, extinction, fossilization) were taken as the median parameter estimates from the results of the Bayesian inference analyses (see below) using the fossilized birth-death model with sampled ancestors. The stratigraphic ranges of fossil taxa other than the new taxon represented by USNM PAL 768729 (see above) are as follows:

The age range of Exostinus serratus describes uncertainty in its stratigraphic distribution, whereas the age ranges of the other taxa refer to the time span between the lowermost and uppermost occurrences. However, both kinds of ranges are treated identically by the software (see below). Due to the differences in the kind of age range, we refrain from analyses of ancestral-descendant relationships of individual taxa in particular trees (cf. Parins-Fukuchi, 2021).

Bayesian inference: The character-taxon matrix was also analyzed using Bayesian inference in MrBayes v3.2.2 (Ronquist et al., 2012). To do so, several multistate characters were recoded so as not to exceed a maximum of five character states per character in MrBayes. Character 55: states 4 and 5 were combined. Character 56: 0 for state 0, 1 for states 1–2, 2 for states 3–4, 3 for states 5–6. Character 64: 0 for states 0-2, 1 for states 3-5, 2 for states 6-8, 3 for states 9-B. Character 147: 0 for states 0–1, 1 for states 2–4, 2 for states 5–7, 3 for states 8–A, 4 for states B–C. Character 148: 0 for states 0–1, 1 for states 2–3, 2 for states 4–5, 3 for states 6–7. Character 217: as for character 147. Character 249: as for character 56. Character 88, states 4–5 were rescored as 3–4 (state 3 was not present in the original dataset).

The analysis (analysis 2) is based on the Mk character model of Lewis (2001). Additive characters were ordered using ctype. Default priors were used. The analyses were allowed to run with 4 chains (1 cold and 3 heated) for 20 Mgen (million generations) and sampled every 1000 gen. Convergence was ascertained using diagnostics obtained from “sump.” Burn-in was set to 25%.

Tip-dating: We also conducted total-evidence dating using the fossilized birth-death (FBD) model (Stadler, 2010) as implemented in MrBayes v3.2.2 to estimate divergence times and determine whether any of the fossil taxa are directly ancestral to any other sampled lineage. Because the foregoing analyses had identified the taxa of interest as pan-xenosaurs, the analysis focused on that group. Elgaria multicarinata was used as the sole outgroup. In the first analysis (3a), fossil taxa were treated as terminal (i.e. no sampled ancestors were allowed), in order to compare the FBD model with the previous analyses. The analysis ran for 20 Mg. In the second analysis (3b), fossil taxa were allowed to be sampled ancestors. Because this analysis produced results at variance with all other analyses, it was allowed to run for 50 Mg, in order to be very certain about convergence. The dataset was the same as used above. We used the relaxed clock model IGR (Lepage et al., 2007). Similar results were achieved using the TK02 model (Thorne and Kishino, 2005), studied intensively by Simões et al. (2020), except that the relative positions of Restes rugosus and Exostinus lancensis were reversed; as this result conflicts with considerable character evidence amassed by Bhullar (2011) and reflected in the MP results, we tentatively prefer the IGR model. Additive characters were ordered as above.

FIG. 1.

Shaded relief map (based on Shuttle Radar Topography Mission 3 arcsecond/90 m digital elevation model) of the Bighorn Basin, Wyoming, showing the location of the 8abc limestone locality.

As there is no evidence for “diversified” sampling of extant taxa (Zhang et al., 2016), we set the sampling strategy prior to “random.” The dataset covers six living species of Xenosaurus. Since Bhullar's (2011) work, several populations have been recognized as evolutionarily distinct or actually elevated to species status, and further species described (Lemos-Espinal et al., 2012; Woolrich-Piña and Smith, 2012; Nieto-Montes de Oca et al., 2017). The most recent genetic study suggests that as many as 18 species could be recognized (Nieto-Montes de Oca et al., 2017). Thus, we set the sampling probability prior to 0.33 (= 6/18).

The tree age prior (i.e., the age of the split between Pan-Xenosaurus and the outgroup, Elgaria multicarinata) was set to an offset exponential distribution with a mean of 75 Ma (mid-Campanian, earliest pan-xenosaur, from Longrich et al., 2012) and maximum of 100 Ma (oldest putative neoanguimorph, from Nydam, 2000).

Priors on species ages were set as uniform distributions on the lower and upper boundaries of the geochronologic range, as given above, except for the new taxon, which we took to have a fixed age of 55 Ma. Accordingly, we adopted the practice of coding the stratigraphic range of fossil species, where known from more than a single occurrence, as a uniform prior between first and last occurrence; this is not necessarily desirable, because it is equivalent to uncertainty in a single date, but alternative approaches have not yet been fully explored for the FBD model (Matzke and Wright, 2016).

Diversity analyses: Graphics and analyses make use of the phytools (Revell, 2012), paleotree (Bapst, 2012) and strap (Bell and Lloyd, 2015) packages for R v.4.0 (R Core Team,  https://www.r-project.org). Raw species diversity curves for continuous data (precise ranges) were calculated using the function taxicDivCont. Species diversity curves are based on time-scaled trees, including cal3 and FBD trees, and calculated using the function phyloDiv or multiDiv. The continuous rather than time-binned versions of these functions were used because the temporal range of most species is fairly precisely known.

Unresolved polytomies in the consensus tree will bias toward higher diversity at older ages, because the participating lineages are extended back in time to the first-documented or first-diverging lineage (e.g., Brocklehurst et al., 2013). To circumvent this problem for FBD BI analyses, we made use of the large samples of trees from the posterior distributions saved in the tree (“.t”) files that are output by MrBayes. The values associated with each branch in the individual trees are in units of expected number of changes given the base rate of the clock, so by dividing these values (imported as edge lengths into the “phylo” objects in R) by the base clock rate (change/unit time) in the associated parameter file (“.p”) for the particular generation and run, branch lengths for these individual tree samples can be derived in units of time. Indeed, that is how MrBayes computes branch lengths in units of time for the consensus (F. Ronquist, personal commun., 2021). We were then able to plot diversity curves (lineage-through-time plots) for each of these trees individually using the function multiDiv.

Whether a particular species is ancestral to another in the phylogeny was assessed based on the unconstrained MP tree under the cal3 model, which distinguishes between budding (when the ancestor continues to live alongside the descendant) and anagenesis (when the ancestor becomes so transformed as to be recognized as a different species, the descendant). Under the FBD model with sampled ancestors (FBD-SA), the probability that a particular species is an ancestor was assessed by randomly sampling 1000 post-burnin trees from the posterior probability distribution and calculating the proportion of times that the branch length of each fossil species was 0; additionally, the output parameter files indicate the average proportion of ancestral species.

Material Examined

Comparative osteological specimens of Anguimorpha used in this study. Shinisauridae: Shinisaurus crocodilurus (MVZ 204291, SMF PH 84, 91, 93, 94); Xenosauridae: Xenosaurus grandis (MVZ 128947, 128948, 137627, 137786, 137798–137790, UMMZ 149608), X. platyceps (UF 45622, 54559); Anguidae: Abronia deppei (UTA R 5646), A. mixteca (UTA R-10277), Anniella pulchra (UF 51810), Celestus costatus (FLMNH 13254), Diploglossus enneagrammus (MVZ 191042), D. monotropis (MCZ 29682), D. pleii (CAS 200840), Elgaria coerulea (CAS 14509), E. multicarinata (YPM HERR 14098), Gerrhonotus infernalis (TMM-M7129, 7525), Mesaspis moreletii (YPM R13518), Ophisaurus ventralis (CAS 74296, 200896); Helodermatidae: Heloderma horridum (SDSNH 59469, YPM HERR 16804, 16820), H. suspectum (CAS 159494, SMF PH 135); Varanidae: Lanthanotus borneensis (SMF PH 336), Varanus exanthematicus (YPM HERR 10812).

Institutional Abbreviations

Description and Comparisons of USNM PAL 768729

Maxilla: The left maxilla is nearly complete, lacking only its posterior end and part of the palatal flange at the anterior end (fig. 2). The facial process is broad, seven tooth spaces in length from its anterior margin to the lacrimal embayment. Its anterior margin curves rapidly dorsally, reaching an inflection point that marks the posterior extent of the external naris on the maxilla, whereupon it extends posterodorsally and medially (fig. 2A, B). The point, marking the extent of the nasal articulation on the maxilla, does not overhang the premaxillary process, as it does in most Anguidae (exceptions: Anniella and Elgaria), Xenosaurus, and Shinisaurus. The inflection point forms a laterally inflected corner, as in Xenosaurus. After reaching its apex, at midlength, the facial process maintains the same height but extends posterolaterally. A small embayment near its dorsal terminus marks the prefrontal articulation. The posterior end of the facial process is more or less vertical but shows a distinct posterior embayment for the lacrimal. The lacrimal articulation is bounded both by this embayment and by a more posterior inflection point in the lateral maxillary wall. The posterior inflection point is absent in extant Xenosaurus and many anguids (Celestus spp., Mesaspis spp., Ophisaurus attenuatus), but is found in other anguids (Elgaria coerulea, Ophisaurus ventralis), Restes rugosus, Heloderma, Eurheloderma gallicum, Lanthanotus borneensis, and Estesia mongoliensis (Norell et al., 1992: fig. 6; K.T.S., personal obs.). Between these boundaries, the dorsal edge of the facial process has an elongate, shallow facet that is slightly broader and deeper posteriorly. Posterior to the lacrimal facet the lateral wall of the maxilla is sharp edged and decays smoothly in height; it is shallowly concave in lateral view and decays so strongly that the jugal groove is partly exposed in lateral view. A row of seven labial foramina (the fifth filled with sediment) is located above the jaw parapet. Dorsal to these foramina the lateral surface of the facial process is moderately but irregularly rugose; epidermal scale boundaries could not be ascertained. A few tiny foramina are scattered among the rugae. The whole facial process is medially folded about a low, oblique axis that is in line anteriorly with the lateral ridge on the premaxillary process and terminates at the posterior projection that marks the boundary between the lacrimal and prefrontal embayments, i.e., at a point continuous with a presumed canthal crest of the prefrontal. Ventral to this axis the facial process is directed laterally; dorsally the process is directed anteriorly, dorsally, and laterally. The folding does not create a distinct ridge. The medial surface of the facial process is mostly smooth. A strong nasolacrimal ridge (Smith and Gauthier, 2013) is essentially absent, except perhaps ventrally near the palatal shelf (where the bone is covered by adhering sediment that is not easily removed).

FIG. 2.

Nearly complete left maxilla of USNM PAL 768729. A, lateral and B, medial views. Medial view also shows the palpebral, which remained attached to the maxilla. C, Close-up of third (preserved) maxillary tooth from rear. Abbreviations: ASAF, anterior superior alveolar foramen; cr.tv., crista transversalis; fac.pr., facial process; j.gr., jugal groove; l.fac., lacrimal facet; lat.r., lateral ridge (= crista lateralis); l.rec., lacrimal recess; n.fac., nasal facet; palp., palpebral; prf.rec., prefrontal recess.

The palatal shelf of the maxilla is dorsoventrally thin (fig. 2B) and mediolaterally wide, projecting medially a considerable distance beyond the bases of the teeth. In this respect it is similar to the palatal shelf of Xenosaurus as well as gerrhonotine, diploglossine, and annielline anguids; it is narrow in Shinisaurus and many Varanus spp., and wide only anterior to the palatine articulation in Lanthanotus borneensis and Heloderma spp. Anterior to the superior alveolar foramen (SAF), into which plunge the superior alveolar branch of V2 and the maxillary artery (Oelrich, 1956), the dorsal surface of the shelf is flat and slopes slightly laterally (unlike in Shinisaurus, where it slopes medially). Anteriorly, the palatal shelf begins to curve medially, following the trend of the lateral margin of the maxilla, but the shelf is broken medially beyond the level of the anterior margin of the facial process. The crista transversalis, extends anteriorly and slightly medially from the anterior base of the facial process, separating the premaxillary portion of the palatal shelf from the posterior portion. The anterior superior alveolar foramen (ASAF), through which passes the superior alveolar nerve and maxillary artery, opens just anterior to the base of facial process and medial to the crista, as in extant Xenosaurus, Shinisaurus (Smith, 2009; Bhullar, 2011) and certain extinct platynotans like Parasaniwa (Gao and Fox, 1996: fig. 33B) and Palaeosaniwa (Gao and Fox, 1996: fig. 38D). The palatine process of the maxilla, the SAF, and the posterior end of the facial process occur at approximately the same anteroposterior level. The anterior end of the palatine process curves sharply away from the trend of the medial margin of the palatal shelf, achieving maximal width in about one tooth space. Its anterior margin is thus more abrupt than in most Xenosaurus grandis (MVZ 137788 comes close), X. platyceps, most anguids (Ophisaurus ventralis and Elgaria coerulea were exceptions), Varanidae, and Helodermatidae. The medial edge of the process is straight and slightly oblique to the trend of the palatal shelf. The process gradually decays in width posteriorly, slightly more rapidly beyond the end of the palatine articulation. The articulation itself is a lenticular, dorsoventrally thin facet, approximately two tooth spaces in length, which faces medially and slightly dorsally and is bounded dorsally and ventrally by small ridges of bone. We infer that the neurovascular structures that enter the SAF divided or separated shortly upon entering the maxilla. One division turned immediately to exit laterally through the ultimate labial foramen, which is visible when viewing the SAF in dorsomedial aspect. Another division passed deep into the maxilla and ran anteriorly; this division was separated from a more dorsal division, which runs immediately below the palatal shelf, by a small sheet of bone. This sheet is just barely visible in dorsal view. The SAF opens posteriorly because the dorsal surface of the maxilla is depressed behind it. This depression forms a trough that is continuous posteriorly with the deepening jugal groove, which widens slightly posteriorly. The SAF and jugal groove are not continuous in most anguids (including gerrhonotines, anguines, diploglossines, and primitive glyptosaurines like Odaxosaurus piger), nor in Xenosaurus spp. or Shinisaurus.

The premaxillary process is not well preserved. It possesses a crista lateralis (Bhullar, 2010). The crista transversalis (Smith, 2006) is considerably taller, at least posteriorly, where it is visible in lateral view. The premaxillary process was weakly excavated between these two ridges. It is broken medial to the crista transversalis.

Dentition: Fifteen maxillary teeth are preserved (fig. 2B), and approximately two more might have been present at the posterior end of the tooth row. There is no evidence of corrosion to the tooth surfaces (cf. Smith et al., 2021). All teeth are simple and unicuspid, unlike those of Xenosaurus (Gauthier, 1982), but they vary slightly in morphology from anterior to posterior. The first tooth is slender, but tooth girth increases rapidly posteriorly; the fourth tooth was probably at near-maximum size for the toothrow, to judge by its alveolus. The shafts of anterior teeth are straight, but their tips are somewhat recurved (especially the first two); posterior to the eighth, the teeth are only slightly recurved. The bases of middle and anterior teeth are essentially parallel sided, and apical tapering is wholly confined to the crown; the last four teeth expand slightly from base to the level of the jaw parapet, whereafter the crowns taper. This flaring of the teeth at midshaft is a similarity shared with Entomophontes spp. (Smith and Gauthier, 2013), especially E. hutchisoni, but not with other extant anguimorphs; some taxa like the anguid Ophisaurus acuminatus, may have strong mesial and distal cutting edges, but they do not flare (Klembara and Čerňanský, 2020). All teeth show some development of mesial and distal carinae, although those on the anterior teeth are slight, scarcely more than longitudinal angulations on the mesial and distal aspects of the teeth. Even by the ninth tooth they are poorly developed. The most extensive, preserved carinae are on the 12th tooth (fig. 2C); even so, tooth cross-sectional shape is weak. The teeth are fairly high crowned, with about 50% of their height protruding below the parapet of the jaw, except at the posterior end of the jaw, where the teeth, in addition to being absolutely shorter, are relatively shorter crowned. The middle teeth of the tooth row have a height of about 1.14 mm (measured vertically from the base of the nutritive foramen to the apex of the crown).

Jugal: The left jugal is well preserved but lacks the ends of its anterior and temporal rami (fig. 3). The exposed lateral surface is sculptured, but unlike on the maxilla, the boundaries of polygonal epidermal scales can be discerned, at least along the orbital margin, as linear depressions that surround raised patches (fig. 3A). Among anguimorphs, a sculptured jugal is also found in Xenosaurus, Shinisaurus, and Heloderma, and Gao and Fox (1996) stress the widening and sculpture of the ascending ramus. The sculptured patches have undulose surfaces and are pierced in places by foramina. A ventral row of five larger foramina occurs along the ventral margin of the suborbital row of scales; a set of five generally smaller, less obviously colinear foramina opened deep to the suborbital scale row. An elongate impression marks where the jugal was in articulation with the posterior process of the maxilla. The articulation surface is strongly demarcated posteriorly, less well so anteriorly. The dorsoventral extent of the overlap decreases posteriorly and has a distinct kink at midlength. The overlap surface ceases to be visible in lateral view about 2 mm anterior to the quadratojugal tubercle. That tubercle is blunt, similar in size to that of Ex. lancensis (Gao and Fox, 1996: fig. 27E); it is (poorly) defined dorsally by a slight concavity of the posterior margin of the bone. The ascending or temporal ramus of the jugal rises at an angle of ∼58° with respect to the suborbital ramus, curving also slightly medially. Its lateral surface is much smoother than that of the suborbital ramus.

FIG. 3.

Right jugal of USNM PAL 768729. A, lateral and B, medial views. Abbreviations: mx.fac., maxillary facet; qj.tub., quadratojugal tubercle.

In medial view the jugal presents a continuous ridge roughly parallel to the orbital margin (fig. 3B). The ventral surface of this ridge on the suborbital ramus is an articulation surface inserting in the jugal groove of the maxilla; the ridge is located below the dorsoventral midpoint of this ramus. This surface continues posteriorly for nearly 1 mm beyond the point where it ceases to be visible in lateral view. Near the posterior end of this surface there is a kink in the medial ridge, which indicates the location of the ectopterygoid articulation. The medial ridge then shows another inflection point, marked by a small, medial tubercle. The posteromedial face of the ascending ramus—posterior to the medial ridge—is deeply excavated, especially ventrally. For most of the preserved extent of the ramus, the ridge is located at the anteroposterior midpoint (not fully compatible with the types of Čerňanský et al., 2014). A small foramen penetrates the jugal in this fossa, well above the ventral margin of the bone.

The orbital face of the jugal is transversely flat anteriorly but becomes slightly convex near the angle of the jugal and finally broadly rounded toward the end of the ascending ramus (fig. 3B). It is relatively narrow for most of its length, achieving maximum mediolateral width at the junction of the two rami. No foramen pierces the orbital face of the jugal on its preserved portion.

Palpebral: The left palpebral is inextricably stuck to the medial surface of the facial process of the left maxilla and is seen primarily in dorsal view (fig. 2B). The lateral edge faces ventrally, as it is preserved, and the anteromedial edge faces more or less posteriorly. In dorsal aspect it has the form of an equilateral triangle; thus, the posteromedial and posterolateral processes are subequal in length, although the latter is distinctly blunter and more robust. The equilateral shape and short, bluntly terminating posterolateral corner of the palpebral are features shared uniquely with Xenosaurus. Most of the element is dorsoventrally thin. However, much of the anteromedial edge is thickened, and viewed end-on there is a shallow longitudinal sulcus or facet. The posteromedial corner, however, like the entire posterior edge, is dorsoventrally thin and sharp. The medial margin is straight and does not show the S-shape typical of extant Xenosaurus (Bhullar, 2011). Like in Xenosaurus, Shinisaurus, and Varanus, but unlike in Anguidae, the posterior margin is concave.

Coronoid: The coronoid, on which various portions of the jaw adductor musculature insert (Haas, 1960), is preserved in articulation with the other post-dentary bones (fig. 4). It is medially concave in dorsal view and straddles the surangular. The tip of the coronoid is a mediolaterally compressed and posteriorly inflected, a feature encountered in North American Ophisaurus (O. attenuatus, some O. ventralis) and Shinisaurus. The tip is completely flat medially, but it bears a ridge laterally that extends ventrally and somewhat anteriorly to form the anterior margin of a fossa. The posterior edge of the tip of the coronoid extends nearly ventrally at first, then posteroventrally at an angle of ∼44° to the vertical, forming a strong ridge on the posteromedial process. A tiny flange extends anteriorly along the prearticular from the ventral end of the posteromedial process. More conspicuous, however, is the strong posterior flange that extends posteriorly over the prearticular and surangular along the entire length of the posteromedial process. This flange is pierced by a tiny foramen at about midheight. The posterior margin of the flange is undulating. This flange also wraps dorsally over the apex of the surangular to form the floor of the fossa; its ventral margin is notched for the anterior surangular foramen (see below).

In medial view the coronoid is deeply concave ventrally between the posteromedial process and the much longer anteromedial process, whose tip is broken. The medial surface of the latter shows a conspicuous facet at midheight whose dorsal margin is roughly horizontal and that marks the medial overlap of the splenial. Ventrally, then, the anteromedial process of the coronoid inserted between the splenial (medially) and the anterior processes of the surangular and prearticular (laterally). On its lateral surface above the level of the facet the anteromedial process articulated on the subdental shelf of the dentary. The coronoid process of the dentary then extends posterodorsally beyond this and inserts in the crotch between the anterolateral and anteromedial processes of the coronoid. The tip of the anterolateral process is broken, but if its dorsal and ventral margins maintained their trends, very little of the process is missing.

Splenial: The splenial, like the dentary, is not preserved, and it seems likely that these elements were dissociated from the postdentary elements prior to burial. However, facets on the remaining bones (fig. 4) permit the reconstruction of the splenial's posterior end. It overlapped the anteromedial process of the coronoid and probably ended shortly before reaching the point where the prearticular dives under the process. From there, its posterior margin extended first ventrally, then arced posteroventrally, inserting between the prearticular and angular. Its posterior-most extent was level with the apex of the coronoid. Its ventral extent is marked by a deep, horizontal sulcus between angular and prearticular, which shallows anteriorly.

FIG. 4.

Partial left mandible of USNM PAL 768729 in medial view. A, Articular region, with broken retroarticular process. B, Middle region, including coronoid and angular. Abbreviations: ang., angular; cb.I, ceratobranchial I; cn., coronoid; part., prearticular; p.mh.f., posterior mylohyoid foramen; ra.pr., retroarticular process; spl.fac., splenial facet.

Angular: This bone forms the ventral margin of the jaw beneath the coronoid (fig. 4). It extends anteriorly to the end of the mandible as it is preserved and would have terminated well anterior of the end of the toothrow. It is pierced by the posterior mylohyoid foramen at a transverse level just anterior to the apex of the coronoid, and it is around this point that the angular achieves maximal medial exposure. Posteriorly, it rotates laterally and is applied almost exclusively to the ventral margin of the mandible. Like the ventral margin itself, the angular is shallowly ventrally concave behind the coronoid process. It has a tabular form, with nearly parallel medial and lateral margins. These margins converge posteriorly, however, and the bone terminates in a sharp point approximately level with the anteroposterior midpoint of the adductor fossa.

Prearticular: The long anterior process of this element runs lateral to the anteromedial and posteromedial processes of the coronoid and is overlapped ventrally by the angular (fig. 4). The prearticular extends as far as the preserved anterior end of the mandible, significantly in front of the posterior margin of the dentary (however defined). It is sutured with the surangular both medially and laterally. The medial suture is interrupted by the adductor fossa, but the anterior portion of the medial convergence of the two bones is visible behind the posteromedial process of the coronoid, unlike in most anguimorphs, except Xenosaurus spp., some gerrhonotines, and the diploglossine Celestus costatus. This suture is also visible between the anteromedial and posteromedial processes of the coronoid; there, it is dorsally concave. Posterior to the tip of the ventromedial process of the coronoid, the prearticular forms a strong, relatively sharp medial shelf. This shelf extends posteriorly for the full length of the adductor fossa and contributes to the strong ventral flatness of the posterior portion of the mandible.

The adductor fossa is dorsoventrally deep but mediolaterally narrow (restricted; fig. 4), as in many anguimorphs (except Shinisaurus). A canal within the prearticular opens into the posterior end of the fossa. This canal is visible at the broken posterior end of the main mandibular section and was once filled by Meckel's cartilage, which also occupied the floor of the adductor fossa. Some porous and friable material of the same color as the bone is found within the canal at its broken posterior end; this material could represent the calcified cartilage.

The prearticular, like the surangular, is fully united posteriorly with the endochondral articular bone (fig. 4). The articular strongly overhangs the medial margin of the prearticular beneath the articular fossa. The prearticular also formed the retroarticular process, which is broken near its base. The foramen chorda tympani, which conveys part of the facial nerve into the mandible, pierces the retroarticular process near its medial edge just behind the articular fossa. The preserved portion of the retroarticular process was flat ventrally but was dorsally concave, with medial and lateral bounding ridges.

Surangular: Anteriorly, the surangular was overlapped dorsally by the coronoid and the coronoid process of the dentary, and laterally by the surangular process of the dentary. The latter, triangular process extended slightly more posteriorly than the former process and was dorsoventrally short. Clear articulation facets on the surangular indicate that the angular notch of the dentary (the space between the surangular and angular processes) was deeper than the surangular notch (the space between the coronoid and surangular processes) and that the angular process of the dentary was the least posteriorly extensive of the three processes that constitute the posterior margin of the dentary. Just beneath the coronoid on the lateral side, and level with its apex, is the anterior surangular foramen. Beneath the ventral concavity of the coronoid on the medial side is a deep concavity in the surangular, a feature shared uniquely with Xenosaurus among extant anguimorphs. The surangular forms the lateral margin of the adductor fossa (fig. 4). Lateral to the superior margin of the bone is a flat, elongate, laterally sloping surface that extends from the suture anteriorly with the coronoid to the edge of the articular fossa. This surface is bounded ventrally by a moderately sharp ridge, below which the surangular is weakly convex and extends ventromedially. The surangular projects medially over the adductor fossa near its anterior end, partially obscuring it in dorsal view and contributing to its “restriction.” Just above this projection is a longitudinal impression that extends for much of the length of the adductor fossa but is best developed near its midpoint; this impression appears to be unique among anguimorphs. The transversely rounded dorsal margin of the surangular becomes wider and dorsally flatter posteriorly. The posteromedial end of the surangular is marked by a strong tubercle.

Articular: This endochondral element is bilobate in dorsal view (fig. 4). The lateral portion is moderately concave in all directions and separated from the medial portion by a mediolaterally broad, longitudinal ridge. The lateral portion is much larger than the medial portion and also deeper and more deeply concave; these facts suggest that the quadrate had a larger and more bulbous lateral lobe than medial lobe. The posterior boundary between the articular and prearticular is more clearly marked posteriorly than the anterior boundary between articular and surangular. However, the surangular projects posteriorly at the point where it meets the ridge divided the lateral and medial portions of the articular. Along the posterior margin of the articular is a small posterior projection that is located lateral to said ridge.

Ceratobranchial I: A short (1.5 mm) section of a slender, hollow, rod-shaped element is preserved in association with the surangular (fig. 4). It probably represents ceratobranchial I (the only ossified element of the squamate hyoid apparatus: Cope, 1892), which unlike the epipterygoid is hollow.

Dorsal vertebra: Based on its proportions and the shape of the synapophysis, the sole preserved vertebra is probably an anterior dorsal vertebra (fig. 5). It measures 2.4 mm in length along the base of the centrum. The bone is generally well preserved, but portions of the postzygapophyses, the right prezygapophysis, and the neural spine are broken (fig. 5A). The dorsal surface of the left prezygapophysis is inclined medially. From its medial anteromedial end at the junction with the neural arch, the margin turns fairly sharply posterolaterally, then curves smoothly posteromedially back toward the neural arch (the slight irregularities in curvature are due to erosion of the edge of the prezygapophysis). Thus, the anterior-most portion of the prezygapophysis lies close to its junction with the neural arch. The transition from the dorsal surface of the prezygapophysis to the neural arch is smooth; however, a distinct surface for articulation of the postzygapophysis on the neural arch (a zygosphene) is not in evidence. On each side of the nearly vertical, inner surface of the neural arch medial to the postzygapophyses, however, is a distinct, dorsoventrally short articular surface that would have contacted the anterior portion of the neural arch medial to the prezygapophysis. That is, the kind of weak, accessory vertebral articulation found in many Squamata (Gauthier et al., 2012, character 468(1)) is found also in this species. The lateral margins of the postzygapophyses are poorly preserved.

The neural arch is subtly inflated between the prezygapophyses, and the neural canal concomitantly slightly expanded. Its anterior-most portion on the midline is broken, but enough remains to show that the neural spine was absent anteriorly but begins to rises abruptly near the level of the posterior end of the prezygapophysis (fig. 5A). The spine arises abruptly. Unfortunately, its dorsal portion is damaged anteriorly. Posteriorly, its dorsal edge is nearly horizontal (fig. 5B). The spine is low by comparison with most extant anguimorphs in the middle and posterior dorsal vertebrae, except the semifossorial taxon Anniella and crevice-dwelling Xenosaurus and other cryptic forms like certain anguines (Čerňanský et al., 2019) and Elgaria multicarinata (San Bernardino County Museum specimen number A500-2327). Posteriorly, the neural spine increases in mediolateral thickness. At its posterodorsal tip, a small divot is excavated that appears natural; similar divots in modern anguimorph vertebrae are filled with small bodies of calcified cartilage. The posterior tip of the spine does not extend beyond the posterior margin of the postzygapophyses, and each half of the posterior margin of the neural arch is only weakly concave, unlike in most anguimorphs.

FIG. 5.

Dorsal vertebra of USNM PAL 768729. A, dorsal, B, left lateral, and C, ventral views. Note the low neural spine. Abbreviations: cond., condyle; cot., cotyle; n.sp., neural spine; poz., postzygapophysis; prz., prezygapophysis; syn., synapophysis.

The neural canal is large, dorsoventrally taller than the centrum, which is somewhat compressed. In lateral aspect the boundary of the cotyle anteriorly and of the centrum exclusive of the condyle posteriorly are slightly anteriorly inclined but primarily vertical. Ventromedially, however, they tend to curve posteriorly (fig. 5B). Thus, the cotyle in ventral view is anteriorly concave. The dorsal edge of the cotyle is also anteriorly concave. The centrum is depressed in comparison with adult Shinisaurus crocodilurus and Xenosaurus grandis. The ventral surface of the centrum is concavoconvex (fig. 5C). It is straight for most of its length in midsagittal section, but a ventral deflection of the ventral portion of the cotyle gives it a concave outline overall. In transverse section, it is concave, slightly more acutely so just posterior to the synapophysis. A tiny subcentral foramen is set in a subtle longitudinal depression on the right side of the centrum, approximately one-fourth of the way posteriorly.

The synapophysis is elliptical, with a slightly more bulbous ventral portion, and is primarily located at a coronal level dorsal to the centrum (fig. 5A). Although damage to the left prezygapophysis makes it difficult to ascertain, it is likely that the synapophysis was only scarcely visible in dorsal view or was obscured entirely by the prezygapophysis.

Scapulocoracoid: Only a small portion of this element is preserved. The scapula and coracoid are indistinguishably fused (fig. 6), suggesting this individual was sexually mature (Maisano, 2002). The glenoid fossa is saddle shaped. The scapular portion is pierced by the scapular foramen along its thickened posterior margin just dorsal to the glenoid fossa. Anterodorsal to the fossa, just above the probable former location of the scapulocoracoid suture, the lateral surface of the scapula shows a broad swelling capped by a tubercle. The posterior margin of the scapulocoracoid fenestra is preserved well above the level of the glenoid fossa. Directly anterior to the middle of the fossa is preserved the thickened posterior margin of the coracoid foramen. An anterior (or primary) coracoid fenestra is not preserved.

Ontogenetic stage and size: The small size of USNM PAL 768729 is remarkable. As noted above, its maxilla measures only 3.3 mm from the anterior base of the facial process to the notch on its posterior margin, and teeth in the middle of the jaw are 1.14 mm tall. Accordingly, the animal represented by USNM PAL 768729 had a SVL of 54.4 mm with 95% C.I. 50.9–58.2 mm (error on the slope; fig. 7). In her careful study of terminal fusions of skeletal elements in lizards, Maisano (2002) did not specifically look at fusion of the postdentary elements, but indistinguishable fusion of the scapula and coracoid suggests the animal was sexually mature (Maisano, 2002). In contrast, the proposed stem-xenosaur Entomophontes incrustatus commonly has teeth >2.2 mm, which implies an SVL >100 mm (using the same model). Extant species of Xenosaurus typically have maximum SVL between 100 and 130 mm (Lemos-Espinal et al., 2012). Thus, the species represented by USNM PAL 768729 was about half the size of extant species of Xenosaurus.

Phylogenetic Relationships

Under MP, three equally most-parsimonious trees were recovered in both the unconstrained analysis (1a) and the analysis with a molecular backbone (1b). In each, USNM PAL 768729 forms a clade together with Entomophontes incrustatus and En. hutchisoni (fig. 8A). Bootstrap support for this clade is 65% in analysis 1a and 70% in analysis 1b. This is a rather remarkable inference considering that USNM PAL 768729 and Entomophontes share only a single element in common: the maxilla. This clade is the basalmost clade of Pan-Xenosaurus, and bootstrap support for a Pan-Xenosaurus so constituted is <50% in analysis 1a and 76% in analysis 1b. Apart from the addition of this basalmost clade, the topology of Pan-Xenosaurus otherwise corresponds to the results of (Bhullar, 2011), with Restes rugosus, Exostinus lancensis, and Ex. serratus successively crownward taxa on the stem of Xenosaurus, supporting the contention that Exostinus is paraphyletic as presently constituted (Bhullar, 2011). Crown Xenosaurus is strongly supported in both analyses (bootstrap 99% in each). Branch length was substantially lower in the unconstrained analysis: 903 vs. 929 steps.

Thirteen unambiguous synapomorphies (under ACCTRAN and DELTRAN) support the monophyly of Pan-Xenosaurus as recognized here: dentigerous arc of premaxilla tall (3: 0 → 1), more vertical lateral edges of rostral body of premaxilla (4: 0 → 1), premaxillary ethmoid canals bony (6: 0 → 1), >2 colinear anterior foramina in main row on premaxilla (7: 0 → 2), at least 3 anterior foramina dorsal to main row on premaxilla (8: 0 → 3), maxillary labial foramina subequal in size (53: 0 → 1), posterior portion of nasal facet of maxilla folded towards the vertical (59: 0 → 1), shape of palpebral a short triangle (109: 0 → 1), foramen in jugal adductor surface (130: 0 → 1), parietal foramen close to frontoparietal suture (152: 0 → 1), Meckel's groove restricted by suprameckelian lip (191: 0 → 1), posterior decline in dentary tooth height reduced (196: 0 → 1), and low neural spines at least in one-third of vertebral column (214: 0 → 1). Three more are present only under DELTRAN, and 56 more only under ACCTRAN. The highly arcuate rostral margin of the premaxilla recognized by Smith and Gauthier (2013) as a potential synapomorphy of Entomophontes and Xenosaurus is not found on this list, because the plesiomorphic condition is present in Exostinus serratus (Bhullar, 2010) and the premaxilla is unknown in Ex. lancensis and Restes rugosus (Bhullar, 2011).

FIG. 6.

Central fragment of left scapulocoracoid of USNM PAL 768729. The scapula and coracoid are indistinguishably fused. Abbreviations: cc.for., coracoid foramen; gl.fos., glenoid fossa; sc-cc.fen., scapulocoracoid fenestra; sc.for., scapular foramen.

Four unambiguous synapomophies support the monophyly of Entomophontes spp. and USNM PAL 768729: relative length of facial process of maxilla rises (55: 3 → 5), relative height of facial process decreases (56: 4 → 6), shafts of dentary teeth increase in diameter posteriorly (197: 0 → 1) (unknown in USNM PAL 768729), strong mesial and distal carinae on cheek teeth (275: 0 → 1). Four more are present only under ACCTRAN.

FIG. 7.

Relation between maxillary tooth length and snout-vent length (SVL) in iguanid lizards (log-log space). Ordinary least squares regression was used to predict SVL from tooth length. Iguanid lizards were preferred to anguimorphs because the broad spectrum of body size covered by available skeletons did not require extrapolation. Red dot represents prediction for USNM PAL 768729.

Three unambiguous autapomorphies of USNM PAL 768729 are: steeper rise of lacrimal recess of maxilla (46: 0 → 1), posteriormost labial foramina of maxilla larger (53: 1 → 0), maxillary osteoderms as low mounds (62: 1 → 2). Five more are present only under DELTRAN.

Ten unambiguous synapomorphies support the monophyly of more crownward Pan-Xenosaurus, including Exostinus spp. and Restes rugosus: bicuspid teeth (1: 0 → 2) (more on this character in the discussion), anterior end of lacrimal recess of maxilla more posterior (45: 1 → 0), maxilla supradental thickening strong (52: 1 → 2), maxillary osteoderms as low mounds (62: 1 → 2), frontals fused (91: 0 → 1), frontal osteoderms variously platelike and domed (97: 0 → 2), palpebral with fused osteoderms (111: 0 → 1), palpebral medial margin with S-curve (113: 0 → 1), lacrimal large (114: 0 → 1), groove anterior to coronoid facet of dentary deep and long (186: 0 → 1). An additional 11 synapomorphies support this clade under ACCTRAN and an additional nine under DELTRAN.

Cal3 found strong evidence for only a single species being ancestral to others (table 1). Exostinus lancensis was interpreted in 90% of the trees in the sample as a direct ancestor (via budding) of more crownward pan-xenosaurs. Two other species, Entomophontes incrustatus and Ex. serratus, were interpreted as ancestors (via budding) in one-fourth to one-fifth of the trees. BI (analysis 2) yielded a tree in which the topology of Pan-Xenosaurus was nearly the same as in MP (fig. 8B). Only within (crown) Xenosaurus was there a loss of resolution. Support for a clade composed of Entomophontes and USNM PAL 768729 was relatively low (58% posterior probability), as it was for Pan-Xenosaurus (51%). Support for Exostinus serratus + Xenosaurus is high, in fact higher than for the monophyly of Xenosaurus. The posterior probability of the clade Pan-Xenosaurus to the exclusion of Entomophontes and USNM PAL 768729 was 0.93.

TABLE 1.

Frequencies with which particular fossil species are interpreted as ancestors (via budding or anagenesis) in the cal3 and FBD with sampled ancestors (FBD-SA) models. The cal3 model was applied to trees consistent with the molecular constraint. The FBD-SA model is based on trees sampled from the posterior probability distribution.

In Bayesian analyses using the fossilized birth-death process, USNM PAL 768729 is inferred to be a member of a monophyletic Pan-Xenosaurus. Nevertheless, the results are illuminating. In the analysis excluding possible ancestors (3a), similar relationships are obtained as for analyses 1 and 2, except that it is unresolved whether Restes rugosus or Exostinus lancensis is more closely related to crownward Pan-Xenosaurus (fig. 8C). If taxa are allowed to be ancestors (analysis 3b), then a result obtains that is at variance with the results of all other analyses. Namely, Ex. lancensis is inferred to be the basalmost known member of Pan-Xenosaurus, while R. rugosus, Entomophontes spp., USNM PAL 768729, and more crownward Pan-Xenosaurus form an unresolved polytomy. Support for this clade of Pan-Xenosaurus exclusive of Ex. lancensis, however, is poor (posterior probability 0.57), and only small changes in the assumptions (assuming R. rugosus has a point occurrence at 57 Ma instead of a range occurrence from 59–57 Ma) are needed to revert the topology to one in which Entomophontes and USNM PAL 768729 form the basalmost clade of Pan-Xenosaurus.

A sister-group relationship between the species represented by USNM PAL 768729 and Entomophontes thus has relatively strong support under MP and that clade is generally inferred under MP and BI as the basalmost known member of Pan-Xenosaurus. Two species, En. incrustatus and En. hutchisoni, have already been described for Entomophontes (Smith and Gauthier 2013). Both species differ strikingly from USNM PAL 768729 in the extent of osteoderm development on the maxilla—in other words, in external appearance. On the basis of the strikingly different appearance we feel justified in naming a new genus as well as species for USNM PAL 768729.

Taxonomy and Relationships

Gao and Fox (1996) challenged the previous contention (Estes, 1964) that Exostinus lancensis has incipiently bicuspid teeth, arguing that any such appearance might result from breakage of the mesial carina of the posterior teeth. Gilmore (1928) did not attribute incipient secondary cusps to this species, nor did Sahni (1972). Examination of the holotype (USNM PAL 10689) supports Estes' (1964) view in the sense that tiny accessory cusps are in fact developed on posterior cheek teeth and are not attributable to breakage (see also Bhullar, 2011). However, these accessory cusps differ from those seen in Restes and Xenosaurus, which are larger and separated by a moderately strong apicobasal groove; this groove is nearly absent in Ex. lancensis. Such an accessory cusp, even incipient, is clearly lacking in Entomophontes spp. and Blutwurstia oliviae.

The teeth of an unnamed xenosaurid described by Gao and Fox (1996: fig. 27H–J) from the Milk River Formation are flared in a manner similar to those of Entomophontes and Blutwurstia oliviae in having trenchant teeth with sharp mesial and distal blades. In light of the more recent descriptions of Entomophontes and B. oliviae, further study of those specimens, in particular the undescribed premaxilla referred to it by the same authors, is desirable. It seems possible that the unnamed species of Gao and Fox (1996) represents part of the same basal clade of Pan-Xenosaurus. If so, the Aquilan (early Campanian, 84–79 Ma) age of the Milk River record would suggest that the divergence of this clade is in the earlier part of the range computed by our cal3 time-scaling (about 95–70 Ma) and FBD-noSA (96–68 Ma) analyses, and that the divergence of Entomophontes and Blutwurstia from other pan-xenosaurs (65–56 Ma) computed by our FBD-SA analysis is much too young.

The form of the teeth, with slightly flaring shafts at mid-height, is a potential synapomorphy of Entomophontes and Blutwurstia oliviae, and other potential synapomorphies were listed above. However, the osteodermal crust on the maxilla of specimens of the latter (like on the other cranial bones known from En. incrustatus) is much better developed in terms of extent and thickness (Smith and Gauthier, 2013). While similar differences in osteoderm development and ornamentation are documented between juvenile and adult representatives of some extant species, such as the anguid Elgaria multicarinata (Bhullar, 2012), terminal fusion of the scapula and coracoid suggests that the holotype of B. oliviae was ontogenetically advanced (see above). Reduction in body size in the anguid lizard Anniella pulchra is associated with reduced osteodermal ossification, a phenomenon attributed to paedomorphosis (Bhullar and Bell, 2008). It is possible that paedomorphosis in B. oliviae is responsible for the lesser ossification of the dermis overlying the cranial bones.

Ghost Lineages, Pseudoextinction, and the Fossilized Birth-Death Model

It has long been recognized that hypotheses of phylogenetic relationships can be incorporated into studies of past diversity, including diversity during mass extinction events (Norell, 1992; Archibald, 1993; Smith, 1994). In the most basic form, tree topology can imply the existence of “ghost lineages” that are unrecorded as fossils. On the other hand, it has also been recognized that ancestral-descendant relationships may play a significant role in the interpretation of extinction and origination rates in the fossil record (Archibald, 1993; 1994). In particular, if one morphospecies becomes so transformed that it is recognized as a new taxonomic species, then the “extinction” of the ancestor is not a true evolutionary event but rather “pseudoextinction.” Archibald (1993) distinguished several models of speciation (different kinds of cladogenesis and anagenesis) and studied their implications for cladistic results. Modern models still distinguish between anagenesis, which does not add to diversity, and cladogenesis, especially budding, which does add to species diversity (e.g., Bapst, 2013; Bapst et al., 2016).

As calculated by Tracer v1.6.0, the proportion of fossils inferred to be ancestors in our analysis had a median of 0.33, that is, two of six fossil species (mean: 0.39, 95% credibility interval 0.17–0.83). From the consensus tree, the two species with the shortest branch lengths are Ex. lancensis (median branch length = 0, mean branch length = 0.053, 95% credibility interval = 0.0–0.37) and R. rugosus (median branch length = 0, mean branch length = 0.35, 95% credibility interval = 0.0–1.57). These two species are the best candidates for direct ancestors according to this model, raising the possibility that the two ghost lineages implied under maximum parsimony are too long (depending on the extent to which the descendant arose before the nominal extinction of the ancestor).

A literal reading of the fossil record (fig. 9A) shows that pan-xenosaur species diversity peaked (with a maximum of two) species in the early Eocene. There was a single known species extant at the K-Pg boundary, Exostinus lancensis, and it survived. A basic time-scaled phylogeny (not shown), however, based on the MP tree, shows that diversity was higher in the Late Cretaceous and Paleocene. This arises because the phylogenetic hypothesis implies that three more lineages first known in the Paleogene extended back into the Cretaceous: the clade Entomophontes spp. + B. oliviae as well as R. rugosus and more crownward Pan-Xenosaurus. Similarly, the three-rate calibrated (cal3) phylogenies also imply a shift of the peak of pan-xenosaur diversity to the latest Cretaceous and Paleocene (fig. 9B), with a median diversity of 4 species at the K-Pg boundary. The FBD-noSA trees similarly show an early diversity peak that encompasses the K-Pg boundary (fig. 9C); a median diversity of 4 lineages (2.5% and 97.5% quantiles: 2–6). This count is reduced for FBD-SA trees to a median of 3 (fig. 9D) in part because some lineages are ancestral to others. Furthermore, both the cal3 and FBD-noSA results suggest divergence times that occurred earlier than the divergences implied by basic time-scaling. The analysis using the TK02 clock model (results not shown) would produce results closer to the FBD-SA analysis.

It has been considered that phylogenetic trees will tend only to increase past taxic diversity, because they only extend lineages back in time (Wagner, 2000). For this reason, the backward temporal shift in the known peak of pan-xenosaur diversity is not very surprising. That this peak now embraces the Cretaceous-Paleogene boundary is of greater interest.

The nadir in species diversity in the late Eocene-Miocene could be explained as an artifact of inadequate sampling. It has been widely recognized (e.g., Estes, 1970; Gauthier, 1982) that taxa of tropical aspect had a greater latitudinal extent in the Eocene, when global climate was warmer and more equitable. Smith (2006, 2009) promulgated a biogeographic model in which taxa adapted to megathermal climates responded to global cooling after the Eocene by retreat or restriction toward the geographic tropics. Although we recognize that one pan-xenosaur species, Exostinus serratus, is known from the early Oligocene, if this clade behaved similarly, then much of its later Cenozoic evolution might be found in the southern United States and Mexico, from which few small reptile fossils have been described.

Paleoecology and K-Pg Survivorship in Pan-Xenosaurus

The low neural spine of the dorsal vertebrae in B. oliviae was noted above as an apomorphy suggesting a relationship with Xenosaurus. It is also a feature of ecomorphologic interest, suggesting a depressed body form or habitus. Dorsal vertebrae are unknown from Entomophontes, but a referred premaxilla (Smith and Gauthier, 2013: fig. 24A, B) has a nasal process projecting posterodorsally at a relatively low angle to the horizontal. While a precise estimate of the angle is difficult because the bones are disarticulated, the angle is lower than for other cryptozoic terrestrial lizards like Diploglossus (e.g., Bochaton et al., 2016). Furthermore, the facial process of the maxilla is also lower in Entomophontes incrustatus (Smith and Gauthier, 2013: fig. 24D), as it is in En. hutchisoni (Smith and Gauthier, 2013: fig. 26A) and B. oliviae (see Diagnosis above). The generally short nasal process in Xenosaurus makes comparison with extant species difficult (see Bhullar, 2011: figs. 9, 10). These features of Entomophontes also point to a depressed body form, or at least a flattened skull profile. Together, they suggest that members of even the basal-most clade of Pan-Xenosaurus may have been crevice-dwellers—and parsimony optimization implies that occupation of this microhabitat is primitive for Pan-Xenosaurus at least as far rootward as that clade.

FIG. 9.

Evolution of species diversity in Pan-Xenosaurus based on various models. Vertical red lines mark the K-Pg boundary. A, Raw species diversity calculated from ranges of known fossil species and no phylogenetic considerations. The peak of two species is in the early Eocene. B, Lineage-through-time plot based on cal3 model of Bapst (2013). The first peak is much higher and encompasses the Cretaceous–Paleogene boundary. C, Lineage-through-time plot based on 100 tree samples from the fossilized birth-death (FBD) analysis without sampled ancestors, showing similar peaks. D, Lineage-through-time plot based on 100 tree samples from the FBD analysis with sampled ancestors, showing similar peaks.

Another feature of ecological interest is size. Members of the basal clade were around 100 mm SVL or less (see above), and similar values obtain for other members of Pan-Xenosaurus. Small lizards tend to be insectivorous rather than carnivorous or herbivorous (Pough, 1973), and the dentition of known members of Pan-Xenosaurus shows no adaptations for herbivory.

These features might have been particularly useful to pan-xenosaurs across the K-Pg boundary because they protected the animals from short- and medium-term consequences of the K-Pg bolide impact. A proposed consequence of the impact is a thermal pulse resulting from the transformation of kinetic energy into heat upon the reentry into the atmosphere of ejecta (Robertson et al., 2004; Schulte et al., 2010). The heat may not have been sufficient to cause woody biomass to ignite, especially wet woody biomass, but it would have a major deleterious influence on nonwoody vegetation. It is potentially related to the so-called fern spike, which has been documented in different regions of the globe and may indicate a pioneer vegetation (Vajda et al., 2001; Schulte et al., 2010). A further consequence of the impact may have been a severe restriction of primary productivity via atmospheric dust and aerosols, although productivity may have recovered within a time scale of decades (Sepúlveda et al., 2009).

We suggest that the higher species survivorship of pan-xenosaurs than other squamate clades is related to their crevice-dwelling habits, facilitated by small size (Longrich et al. 2012). Crevice-dwelling habits could have provided shelter from the initial thermal pulse, while insectivory made them less reliant on photosynthetic energy pathways. Similarly, aquatic vertebrates, except for elasmobranchs, are well documented to have a much higher survivorship across the K-Pg boundary (Archibald and Bryant, 1990; Fastovsky and Sheehan, 2005), possibly because the high heat capacity of water protected them from the thermal pulse, while their food chains were less dependent on photosynthesis (Sheehan and Hansen, 1986; Robertson et al., 2004).